Measurement Method For Standing Tree Attributes Based On Depth Map

The DBH(Diameter at Breast Heigh),TH(Tree Height)and crown width of standing trees are important measurement factors in the forest resource inventory,and are important basis for assessing forest land accumulation and forest growth.At present,the measurement method of standing tree factors still has shortcomings such as low universality and low measurement efficiency.Therefore,this paper focuses on a depth map-based measurement method of standing trees.This method uses consumer-grade portable equipment(smartphone,Kinect)as tools to collect standing tree images in natural scenes,uses three-dimensional measurement technology to explore the method of three-dimensional reconstruction of standing tree in natural scenes,and calculates the DBH,TH and crown width of standing tree through three-dimensional model of standing tree.The main contents of this paper include:(1)A method for generating depth maps of standing trees based on short video images.In this method,foreground and background of the first frame of the short video are marked through the Graph Cut algorithm,and the sub-pixel corner points are extracted.Tracking and matching the obtained sub-pixel corner points,using the bundle adjustment to estimate the camera parameters and the spatial position of the feature points.Correcting the distortion of each frame of the video based on the distortion parameters and implementing dense stereo matching combined with the plane scanning method.Then fine depth map is obtained by denoising.(2)A method for 3D reconstruction of standing trees based on depth maps.According to the transformation relationship between image coordinate system and object coordinate system,the depth map is transformed into point cloud image.The target tree point cloud data is segmented from the original point cloud images by using the through filters.Based on the color region growing segmentation method,the standing tree point cloud is divided into two parts: the main trunk point cloud and the crown point cloud.Through SDRSAC-NDT(Semidefinite-Based Randomized Approach for Robust Point Cloud Registration without Correspondences-Normal Distribution Transform)method,the main tree point cloud is registered,and the rotation-translation matrix are obtained.And then,to reconstruct the three-dimensional model of the standing tree,the leaf point cloud is filled.(3)Measuring method of standing tree factors based on three-dimensional model.In the end,based on the three-dimensional model of the standing tree,the key point information of the standing tree is extracted,and establishing the spatial geometric relationship to measure the DBH,TH and crown width of the standing tree.This paper proposes a measurement method of the DBH,TH and crown width based on the consumer-grade portable equipment.Through the verification in the natural environment,this method can effectively generate a fine tree depth map,and the relative error of tree height measured by this method is less than 2.89%,the relative error of crown width is less than 7.841%,and the absolute error of tree diameter is less than 0.28 cm when the DBH is no more than 20 cm.Therefore,the method proposed in this paper can meet the accuracy requirements of national forest resources inventory.